Deep Learning Based Foreign Object Debris (FOD) Detection on Runway

An improved method for detecting Foreign Object Debris (FOD) in a runway environment is presented in this paper. The method involves implementing detection model based on deep learning algorithm which can classify and detect object with high accuracy i.e mean Average Precision (mAP). Regular method like radar technology and optical imaging system are used more frequently but these are sensitive to weather and this affects the accuracy. To avoid this Computer Vision (CV) and Machine Learning (ML) are being used. In this work, use of YOLO model has been done and the dataset taken is named FOD in Airports (FOD-A) which is publicly available and contains 31 object categories (nuts, bolts, washers, safety wires, etc.) and over 30,000 object instances. Deep learning model YOLO v5 and YOLO v8 are used for FOD detection and results are compared. In the study, it is analysed that, the average model accuracy, specifically mAP50 and mAP50-95, improved by 0.163% and 7.714%, respectively. The study brings the efficacy of YOLO v8. Here, YOLO v8 ensured better results with mAP50 of 99.022% and mAP50-95 of 88.354% for FOD detection.